A version of this essay was originally published at Tech.pinions, a website dedicated to informed opinions, insight and perspective on the tech industry.
We’ve now had two major developer events in a row where chat bots were a significant theme, with both Microsoft’s Build and now Facebook’s F8 focusing on this rapidly emerging new form of interaction with companies and brands. With two such big names behind the trend, it’s easy to get caught up in the hype and enthusiasm these companies obviously share for the technology. But it’s important to stay grounded as we evaluate chat bots as a potential successor to today’s app model.
The first thing to note is that Facebook and Microsoft have strong incentives to pursue the bot vision. Both companies failed to make a meaningful dent in the mobile operating system battle and, as such, find themselves in secondary roles as makers of apps that run on other people’s platforms. This shuts them out of many of the opportunities associated with owning a mobile operating system, and puts them perennially in a secondary position, having to work around the limitations placed on third-party apps and the inherent disadvantages they face relative to pre-installed applications. So it’s not surprising both companies are now embracing what — in at least some visions of the future — promises to be the replacement for mobile apps. But it’s important to keep these incentives in mind in evaluating their claims about the potential of bots — Facebook and Microsoft have a massive vested interest in seeing this trend succeed.
Culture matters, too
Many users will have made existing investments of time (and in some cases, money) in mobile apps for some of the same interactions bots will now offer.
The other thing that matters enormously is culture. From conversations I’ve had recently with proponents of the chat bot model as a successor to mobile apps, it’s clear that their arguments are strongest in cultures where messaging has become the dominant model of interacting with the world, for everything from intimate conversations with significant others to confirmation emails for food orders or plane tickets. However, this culture isn’t universal by any means — it’s far stronger in some parts of the world such as Asia, and probably weakest in markets like the U.S. The version of the future in which bots dominate is highly dependent on a present where messaging dominates, and that simply isn’t the case in all markets today. The app model dominates in many markets today, and such adoption of bots requires a significant mind shift by users, and a break with what are now fairly well-formed habits built around apps. There are, of course, also major generational differences around messaging behaviors.
Where bots work
The fact is, chat bots work well for certain interactions that have specific characteristics. Let’s discuss each of those briefly.
Interactions must be quick. If a conversation with a bot takes more than a handful of messages, it almost certainly would have been quicker and more efficient using a touch-based Web or app interface. The fewer the number of back-and-forth exchanges, the better suited a task is to a bot interface. Typing out endless responses to questions rather than simply pressing buttons makes the model unworkable.
Interactions must be simple. If the user has too many options to choose from, the bot model becomes unworkable. Onstage at F8 on Tuesday, Messenger head David Marcus demonstrated the process of buying a pair of sneakers from an online store through a bot. But the bot only presented him with five possible options to choose from. For the sake of simplicity, that’s where the demo ended. But the reality is, the odds that one of these five pairs of shoes would be the right one in real life are low. I’ve tested the same bot, and there are ways to get more options beyond the five, but they come in additional batches of five only, based on a request to see more. For anyone who’s browsed an online store on a website or in an app with a massive grid of options, this five-at-a-time experience is likely to be frustrating. It only works well when there are limited choices available and the user can quickly burrow down to the right one.
Context is available to the bot. The bot uses context in the same way an app does — there’s a history between the customer and the company, cues such as location are available, and so on. Facebook’s “weather cat” bot Poncho seemed to fall flat in the hours following its launch by prompting users to provide a location in text, many of which it didn’t recognize. It did also provide the option to send the current location through the standard Messenger interface, but many users seem not to have understood this was possible. But reducing the load on the user by pulling in all relevant cues and context will be critical to making bot interactions efficient. Tying users to existing customer profiles is clearly part of this, as well, and Facebook has a limited number-matching system for doing this.
Users maintain control. Users have to maintain control in order to give bots permission to enter what can otherwise be a very personal space — their messaging app. There were several failures on this front in the first day of Facebook’s new bots. News apps like CNN took any message as permission to spam the user indefinitely with breaking news without asking for an opt-in. Many of the bots appear to send news messages (and thus trigger OS-level notifications) by default until the bot is blocked, which is awkward terminology, to say the least, and probably overly heavy-handed, except for the fact there seems no other obvious way to stop them from sending messages.
Users haven’t made existing investments in apps. Perhaps the biggest challenge, though, is many users will have made existing investments of time (and in some cases, money) in mobile apps for some of the same interactions bots will now offer. Why would someone who’s made that investment in a dedicated app switch to using a bot? The biggest opportunity here may well lie in the same space Google’s app-streaming model is intended to target — new or temporary interactions where the user hasn’t made an investment and/or doesn’t want to. Examples would be one-off purchases from a new store, booking a restaurant in a new city while on vacation or a business trip, or checking transit times while your car is in the shop for a day.
A “sometimes” solution
Bots are — for today, at least — a “sometimes solution” for interactions with companies and brands.
Sesame Street’s Cookie Monster, in his more responsible modern incarnation, now refers to his beloved cookies as “a sometimes food,” in contrast to fruits and vegetables and other foods that can be eaten more regularly. What’s clear from everything I’ve outlined above is bots are — for today, at least — a “sometimes solution” for interactions with companies and brands. There will be some interactions for which they work, and work well, but many others for which they’re too cumbersome, too inefficient, underperforming and simply require too much of a learning curve to be useful or pleasant.
The challenge for Facebook, Microsoft and every other company pitching the bot vision is to refine that vision to meet the specific challenges for which it works, and execute well on those, rather than to continue to spread the idea bots will replace apps for essentially all interactions. That won’t happen, at least anytime soon, and these companies are doing themselves, their users and their paying customers a disservice when they pretend it will.
Jan Dawson is founder and chief analyst at Jackdaw, a technology research and consulting firm focused on the confluence of consumer devices, software, services and connectivity. During his 13 years as a technology analyst, Dawson has covered everything from DSL to LTE, and from policy and regulation to smartphones and tablets. Prior to founding Jackdaw, Dawson worked at Ovum for a number of years, most recently as chief telecoms analyst, responsible for Ovum’s telecoms research agenda globally. Reach him @jandawson.
This article originally appeared on Recode.net.